Achievement and behaviour in undergraduate mathematics: personality is a better predictor than gender

Abstract

We investigated two factors that predict students' achievement and behaviour in undergraduate mathematics: gender and personality. We found that gender predicted students' achievement and behaviour when considered in isolation, but ceased to be predictive when personality profiles were taken into account. Furthermore, personality accounted for significantly more variance in undergraduates' achievement and behaviour than did gender, but the converse was not the case. We therefore argue that personality provides the more productive lens through which to understand the behaviour of undergraduate mathematics students. We relate this finding to recent research emphasising gender differences in mathematics education, and suggest that researchers wishing to promote equity in participation at and beyond the undergraduate level should consider shifting their focus to individual differences in personality.

Keywords:

• gender
• personality
• undergraduate

Acknowledgements

This work was partially funded by a grant from the National HE STEM programme (to SK) and by a Royal Society Worshipful Company of Actuaries Fellowship (to MI).

Notes

1. A re-analysis of their data indicated that on average women mathematics students attended 62% of their lectures and men attended 49%, t(206.8) = 3.484, p = .001, d = 0.459.

2. Our dataset is available at http://dx.doi.org/10.6084/m9.figshare.865640.

3. Prior to conducting the regression and ANOVA analyses we checked our predictors for collinearity. As would be expected (the five personality variables are, by definition, orthogonal), there was no evidence of this: all VIF values were well within the acceptable range (all < 2).

4. For seven participants we had missing data for one or two variables (e.g., because a participant did not complete every questionnaire item). A total of 8 values (0.69% of the dataset) were missing; such cases were dealt with using pairwise deletion.

5. This ensured minimal ΔR² values for the most predictive of the personality variables.